Saul Smilansky: Libertarian free will is impossible, Pt. 2

Shortly after posting my first critique of Smilansky's disproof of Libertarian Free Will, I ran across the book version of his argument. Google Books is awfully handy sometimes.

The argument in the book, as one might expect, provides more detail than the one he presents in the paper I critiqued. I should note that Smilansky emphasizes that his treatment of the issue in the book is also deliberately brief.

In this more developed account, Smilansky relies on the Principle of Sole Attribution, and he offers the following expression of the PSA:

Any feature F due to which a person deserves something S in the libertarian free will-dependent sense must, in the normatively relevant respects, be solely attributable to the person or to the pertinent aspect A of the person.

At first blush, the PSA seems like a human responsibility equivalent to the Principle of Sufficient Reason. The latter, I think, qualifies as corollary of causal determinism. As for the PSA, Smilansky states that "solely attributable" constitutes the key.

For now, it suffices to define it negatively, in that no one or nothing beyond the person's authorship and control can be allowed, in the normatively relevant way, to have brought forth F, if the latter is to be a source for desert in the libertarian sense.

As there may be hidden assumptions in the above, I would decline to agree pending additional information. If I naturally like the taste of salt and cannot help liking salt, I do not see why that would take away my responsibility for eating salty foodstuffs so long as other conditions for libertarian free will were met.

It turns out, at least apparently, that Smilansky obtains his view from libertarian advocates such as C. A. Campbell:

Campbell manifestly attempts to meet something like the criteria laid down by the PSA. He takes a limited aspect of a person, focused on a narrow concept of desire, and calls it 'character.' He then invokes the 'self' as a latent power, which in certain situations can cause us to act contrary to our 'character' (i.e. contrary to our strongest desire).

I call it a mistake to suppose that the self does not cause outcomes corresponding to the strongest desire. The point should be that the self acts as the cause significantly regardless of the strongest desire, as illustrated by modeling LFW for a single act such as consuming potato chips.

Suppose "Homer" is repeatedly put through a trial under conditions x where he may eat potato chips or not eat potato chips. It isn't that Homer eats the potato chips unless the self intervenes. It is that either option is caused by the self without in turn being absolutely caused by something other than the self. So if Homer eats the potato chips 99 times out of 100 it still does not follow that the outcome was caused by the desire to eat potato chips as opposed to the self.

My illustration points up the difficulty of separating the character from the particular leaning of the will (the self) in making the actual decision. I think Smilansky tends to view the character in terms of causal determinism, hence the conclusion that the self would have overruled the character in order to avoid eating potato chips. That is, eating potato chips would have been causally determined. Doing other than eating potato chips would be the indeterministic self. I think Smilansky fails to keep to indeterminism in forming his critique. If an event happens only 99 times out of 100 under identical conditions then all 100 trials are indeterministic, not just one of them. Determinism by definition requires all 100 trials to reach the same outcome. Claiming any number short of that as deterministic creates a contradiction.

The problem comes out clearly in Smilansky's book when he turns to E. Walter for explanation:

"Even if it were plausible to introduce a 'self' to explain behaviour, we would say (a) that the self's decisions are determined by the self's attributes and (b) the character of the substance 'self' is (at least partially) determined at the moment of its inception in accord with whatever laws relate to the nativity of such recondite beings ... We would have no reasonable explanation for how the self gets to be the way it is unless it derives its character potential at birth."
(ellipsis reflects my edit)

Beyond the fact that Walter looks inclined to impose determinism as the explanation for indeterminism, he apparently waves off probabilistic explanations, such as the one I offered above, with no consideration at all.

From there, Smilansky invokes the whole of a Galen Strawson argument that I have tried and found wanting.

Smilansky subsequently summarizes the argument in two parts and surprises me with his follow up:

At this stage it might be said that my argument for the incoherence of a worthwhile sense of self is begging the question; I introduce the requirement of determinism and then am "surprised" to find that libertarian free will cannot meet it.

Bingo! But Smilansky tries to argue that the requirement of determinism is in turn required by the argument of the LFW advocate: "(W)e have not required from the libertarian more than she must require from herself: an account must be given, as to which non-arbitrary factor brought about the decision or action, and why."

Not surprisingly, that argument is circular in its turn.

Smilansky seems to overlook the fact that the LFW paradigm explicitly and intentionally fixes the self as the type of arbiter that he apparently finds unsuitable. The decisions of an arbiter are, in fact, unavoidably arbitrary in an essential sense. Arbiters arbitrate.

Consider an example. Parson Brown realizes that it would be wrong to trample his neighbor's petunias. Yet he feels as though he would enjoy trampling the petunias just the same. Given that Brown knows it is wrong to trample the Petunias, how can he be morally responsible if he tramples them? That is the question Smilanksy seems to offer. Yes, it seems arbitrary for Brown to trample the Petunias when he knows better. But it is precisely his knowledge that stepping on the flowers is wrong that gives him his moral responsibility for his behavior should he trample them in fact.

Doubtless it may be argued that the desire to trample the petunias was too great to resist--but once we take that route we're right back begging the question of libertarian action. Where indeterminism applies in this specific instance, a temptation too great to resist is an impossibility.

Smilansky is correct that the self must have some sort of nature or character. However, he appears to err in the assumption that said nature is necessarily deterministic (at least to achieve personal responsibility), and in the assumption that the self could not control decisions that alter said nature.

*****


Smilansky takes up objections based on indeterminism in the next section of his book, so I will have reason to post a third part to this series.

Saul Smilansky: Libertarian free will is impossible, Pt 1
Saul Smilansky: Libertarian free will is impossible, Pt. 3

Saul Smilansky: Libertarian free will is impossible

On a lark, I went hunting for a claim that libertarian free will is impossible. And not just any claim. I am after the best arguments I can find. I found the claim readily enough from Saul Smilansky. Unfortunately, his argument owes a huge debt to that of Galen Stawson, which I have already addressed.

Still, there is no great harm in repeating the exercise. It always remains possible that some element in the explanation will either add to the earlier argument or add to my perception of that earlier argument.

So without further ado, Smilansky's brief presentation of the Strawsonesque argument:

The reason why libertarian free will is impossible, in a nutshell, is that the conditions required by an ethically satisfying sense of libertarian free will, which would give us anything beyond sophisticated formulations of compatibilism, are self-contradictory, hence cannot be met.

That sets forth Smilanksy's position, so he continues with the rationale:

This is so irrespective of determinism or causality. Attributing moral worth to a person for her action requires that it follow from what she is, morally. The action cannot be produced by a random occurrence and count morally. We might think that two different things can follow equally from a person, but which one does, say, a decision to steal or not to steal, again cannot be random but needs to follow from what she is, morally. But what a person is, morally, cannot be under her control. We might think that such control is possible if she creates herself, but then it is the early self that creates a later self, leading to vicious infinite regress.

The first key point:

"Attributing moral worth to a person for her action requires that it follow from what she is, morally."

I believe the LFW advocate should agree with this proposition, with the caveat that what a person "is" need not be the result of causal determinism. Smilanksy suggests that the argument hold regardless of "determinism or causality"--an odd thing to mention since the LFW position is an incompatibilist one. After all, if determinism is true, then we have reason to expect that LFW is impossible.

The second key point:

"The action cannot be produced by a random occurrence and count morally."

As with Strawson's argument, I detect a concept that requires some explanation. What is a "random" occurrence? A LFW model would (ironically) predict unpredictable outcomes except perhaps in terms of probabilities.

For example, suppose we were to model the decision to steal or not to steal according to LFW. LFW would suggest that an infinite number of trials under identical conditions would produce at least one trial unlike the others. If we have 999,999 trials where the subject steals and a single trial where the subject refrains from stealing, we can claim some element of "chance" in the outcome. But is the outcome random any more than the other 999,999?

On this point, I attempt to clarify the issue by distinguishing between random/chance cause and random/chance outcome. If Smilansky does not detect randomness by mapping outcomes then he appears to have it tucked into his assumptions--or else I should expect an explanation as to how he determined its presence.

Developed further, this idea turns into the notion that who the subject is may well amount to "an individual who steals under conditions x 999,999 times out of a million." This appears to satisfy the qualification that the cause of the outcomes was not random as well as the stipulation that the individual acted according to "what she is."

On to the third key point:

"But what a person is, morally, cannot be under her control."

We might well ask how that follows, and Smilansky obliges, albeit in brief:

We might think that such control is possible if she creates herself, but then it is the early self that creates a later self, leading to vicious infinite regress.

I think that both Strawson and Smilansky overlook the fairly elegant libertarian solution to the problem. Libertarians do not regard individuals as responsible for their own ultimate existence. However, libertarian models appear to allow an individual to create what a person is, morally, subsequent to being created.

For example, suppose we have "Alex," a newly created entity with a non-deterministic will. Alex has a 50/50 chance of stealing/not stealing under conditions x by default. That is how Alex was created, and we do not hold Alex responsible for that state. In addition, Alex has a 50/50 chance of choosing a morally superior lifestyle. If Alex chooses moral improvement, it increases the not stealing probability by one percentage point. The result? Alex is in at least partial control of having a 49/51 chance of stealing/not stealing, and in full control of having the 49/51 figure instead of 50/50. If this decision does not reflect personal responsibility, then at least the subsequent decision not to steal under conditions x does reflect personal responsibility.

If I'm correct, then Smilansky's third statement is simply false, and his argument fails.

I imagine that figures like Smilansky and Strawson do not view things this way because they have a tendency to model things in terms of determinism. There is a determined outcome and anything else has to be random and therefore beyond individual responsibility.


Saul Smilansky: Libertarian free will is impossible Pt. 2
Saul Smilansky: Libertarian free will is impossible Pt. 3

Galen Strawson out of luck

Galen Strawson, philosophy professor at Oxford, wrote an essay on free will and determinism in 1998 titled “Luck Swallows Everything.” The essay offers a skeptical account of libertarian free will, including a semi-formal critique. For that critique, Strawson channels LFW skeptics, calling them “Pessimists” and sets forth a nine-part objection to LFW.

The first of the nine consists of the proposition that people act because of who they are at any given moment of decision. This first would-be premise immediately reveals a nagging problem for Strawson’s essay: His writing carries a bit too much ambiguity. If “who they are” consists of the person prior to the decision, which seems likely given the words used, then Strawson may be guilty of assuming causal determinism from the outset with respect to human decisions. If “who they are” consists of the person and their decision then the statement is acceptable, but not particularly useful on its face since it does not distinguish between differing models of decision making.

The second of the nine relates the supposed LFW claim that “ultimate responsibility” is required for moral responsibility. Again, Strawson writes somewhat ambiguously, referring to “at least in certain respects” but those certain respects are not specified. That may be the fault of LFW advocates, though philosopher Robert Kane does offer a fairly detailed description.

The idea is this: to be ultimately responsible for an action, an agent must be responsible for anything that is a sufficient reason (condition, cause, or motive) for the occurrence of the action.
(Kane, “The Oxford Handbook of Free Will,” p. 407)

The third of the nine is set as a response to the second. The Pessimist asserts that one “can’t be ultimately responsible for the way you are in any respect at all” and therefore not ultimately responsible for what one does. Note that the Pessimist has drifted from Kane’s description by requiring the agent to be ultimately responsible for anything of sufficient reason for the occurrence of the action instead of simply responsible.

The fourth of the nine attempts to justify the third. Rather than justifying any difference between Kane’s version and the Pessimist version of the requirement, however, number four simply expands on what may be a straw man: “To be ultimately responsible for the way you are, you must have somehow intentionally brought it about that you are the way you are.” Kane’s position differs from this statement, saying that the agent must be merely "responsible" at least in part. Who is right? If Kane is correct then Strawson’s Pessimists have set upon a straw man, in effect.

With the fifth of the nine steps, Strawson’s Pessimists launch into an example. We are to suppose the existence of an agent which has brought about a mental state Z for which it is ultimately responsible.

With six of nine they object to the aforementioned agent. To be ultimately responsible for state Z, “You must already have had a certain mental nature Y, in the light of which you brought it about that you now have Z,” adding that the lack of a Y leading to Z indicates no mental state and thus no means of having responsibility for Z. Proposition six is unobjectionable on its face. With step seven, though, the Pessimists again differ with Kane’s model.

Step seven claims that in order have ultimate responsibility for state Z, the agent must also have ultimate responsibility for Y. The Pessimists obviously intend to suggest the problem of an infinite regress.

Steps eight and nine confirm the prediction:
So
(8) You must have brought it about that you had Y.
But then
(9) you must have existed already with a prior nature, X, in the light of which you brought it about that you had Y, in the light of which you brought it about that you now have Z.

Step eight as it stands is inoffensive, since it makes no questionable presumption about “ultimate responsibility.”Step nine, however, implicitly refers back to step six. Whereas in Kane’s scenario having prior nature X may be based on causal determinism, Strawson’s framework has stipulated that one must be somehow responsible for prior nature X.

Note how Scott Hall’s paraphrase of Kane’s model contrasts with Strawson’s modeling of LFW:

One way to understand the notion of UR is to think of it in terms of control. In order to be ultimately responsible for our actions, we must be in ultimate control of our actions. To have ultimate control over our actions means that we must have at least some control over the causes that are sufficient to move us to action, or must have had control over some action in the past that is now sufficient to move us to action. We need not have control over every action in order to be responsible. Suppose, for example, someone makes a decision and she is responsible for that decision. Assume that she had at least some control over the sufficient causes that moved her to action. Suppose further that making that decision is sufficient to cause the person to make some future decision. She would still be responsible for the future decision because she was responsible for the sufficient causes of that future decision.
("Kane, Alternate Possibilities, and Ultimate Responsibility")

Hall points out that no causally determined conditions prior to the conscious action of a given actor can manifest a state of control on the part of that actor. Thus, a causally determined universe CDUa will be sufficient to cause later state CDUz. The later state is predictable based on CDUa regardless of control. Kane’s model opens up the possibility of a Ua leading to a Uz where agents possessing control provide the reasons sufficient to lead to Uz where otherwise Uz would not occur. In the latter model, contrary to the first, the initial state never provides the ability to predict the later state.

Hall’s expression of Kane’s concept helps explain where Strawson goes awry, for Strawson never indicates an awareness of the link between control and sufficient reason. For Strawson, the LFW model must account for the actor’s initial state where the actor controls the initial state. Strawson does not justify his implicit stipulation with respect to a model like Kane’s.

Strawson goes on to assert the impossibility of a “self-origination” capable of supporting LFW, then goes on to present his preceding argument in “more natural” terms.

(A) One is the way one is, initially, as a result of heredity and early experience.

Indeed, and we can even discount experience if it implies mental states.

(B) These are clearly things for which one cannot be held to be in any way responsible (this might not be true if there were reincarnation, but this would just shift the problem backwards).

Again correct, though of course Kane’s model implies nothing different if understood without the intervention of straw men.

(C) One cannot at any later stage of one’s life hope to accede to ultimate responsibility for the way one is by trying to change the way one already is as a result of heredity and experience. For one may well try to change oneself,

Strawson’s (C) relies on a change in the definition of “ultimate responsibility” as described by Kane. As a result, Strawson can hardly be taken to refute Kane. For Kane, the non-determined decisions of an agent with control may provide sufficient reason for later mental states.

but (D) both the particular way in which one is moved to try to change oneself, and the degree of one’s success in one’s attempt at change, will be determined by how one already is as a result of heredity and experience.

Much more so than Strawson’s earlier expression of his argument, (D) implies causal determinism.

Consider again Hall’s expression of the Kane model in terms of control. If we make indeterminism a part of the scenario—and given that Kane is an explicit indeterminist we can hardly do otherwise without begging the question—it does not follow that the particular way one tries to change oneself nor the degree of success for the attempt to change oneself follow (causally) from the initial mental state based on heredity and experience. If they merely follow (indeterministically) in conjunction with the control of the acting agent then Kane appears to have produced something akin to the “certain mental respects” that Strawson has declared impossible.

And (E) any further changes that one can bring about only after one has brought about certain initial changes will in turn be determined, via the initial changes, by heredity and previous experience.

With (E), Strawson just makes more explicit his assumption of causal determinism. One simply doesn’t get from Ua to Uz based on Ua without the assumption of causal determinism. Even taking “previous experience” to refer to past mental states of the LFW agent (in an attempt to read Strawson charitably) fails to rehabilitate his claims, since wherever the past mental states were under the free control of the agent we again have a manifestation of a self-caused mental state “at least in certain respects.” With the introduction of indeterminism, it simply doesn’t follow that the subsequent mental states are sufficiently caused by the prior causally determined (as implied by Strawson) mental states. Strawson acknowledges this in (F) and (G).

(F) This may not be the whole story, for it may be that some changes in the way one is are traceable to the influence of indeterministic or random factors.

But (G) it is absurd to suppose that indeterministic or random factors, for which one is ex hypothesi in no way responsible, can in themselves contribute to one’s being truly or ultimately responsible for how one is.

Note what Strawson attempts with (G): He takes the traditional assertion of LFW, that being able to do otherwise is a key to personal responsibility, and in effect asserts exactly the reverse. Let us cheerfully admit that a “random” influence would be counter to personal responsibility. We need to remember that a “random” outcome is exactly what LFW would predict, since it is simply the converse of a causally determined outcome. It appears that Strawson commits a fallacy of ambiguity, here, for he is not entitled—at least not on the basis of his explicit argumentation—to categorize intentioned “random” outcomes along with those “influenced by random factors” where the latter cannot be the responsibility of a conscious actor by definition. In short, Strawson needs to explain that particular hypothesis in a way that justifies his assertion.

With his two arguments complete, Strawson moves to summarize:

The claim, then, is not that people cannot change the way they are. They can, in certain respects (which tend to be exaggerated by North Americans and underestimated, perhaps, by members of other cultures). The claim is only that people cannot be supposed to change themselves in such a way as to be or become ultimately responsible for the way they are, and hence for their actions. One can put the point by saying that in the final analysis the way you are is, in every last detail, a matter of luck - good or bad.

As Strawson has constructed an argument filled with loopholes born of dubious assumptions, it seems to fair to suppose that he has run out of luck.

By way of explanation, if we can simply be lucky enough that our indeterministic desires consistently correlate to desired outcomes, then luck has ceased to have meaning. "Luck" transforms into a tautology explaining nothing if it supposedly explains anything and everything.